Today's fiber optic transport systems are evolving in many ways. They are achieving longer transmission distances between electrical regeneration points through the introduction of coherent detection and digital signal processing (DSP) technologies. Today's systems have evolved from point-to-point systems to optical mesh deployments through the introduction of reconfigurable optical add drop multiplexer (ROADM) technologies. These changes were first applied to terrestrial transmission systems. Recently, these technologies have been introduced into submarine systems which have traditionally required specific transponder technologies because of the long distances between electrical regeneration points.
These changes present opportunities and challenges. The opportunity is to use the flexibility of the ROADM and the DSP to allow the path of the signal to be switched in the optical domain. The challenge is to maintain post forward error correction (FEC) error-free performance for all other signals in the transmission system while making these changes.
A photonic line system is a concatenation of optical amplifiers. For example, these may be a combination of Raman Amplifiers and Erbium Doped Fiber Amplifiers (EDFAs). The performance of channels transmitted through the line system is optimal when the powers of the channels are controlled to balance noise and non-linear effects in the fiber. The effect of noise added by the optical amplifiers is minimized by increasing the power of the channels. However, deleterious non-linear effects are stronger at higher optical powers. Therefore, there is an optimal power level for any individual channel. To complicate this further, there may be an arbitrary number of these channels present in a dense wavelength division multiplexing (DWDM) transmission system. In addition, the gain shape and noise performance across wavelength of the amplifiers is affected by the spectral loading of the system, i.e., the population of other channels, the wavelengths they occupy, and their optical powers. Since channels in an active system may be added or dropped for a variety of reasons, the gain shape and noise performance changes in an unpredictable manner. Non-linear effects are also affected by the spectral loading in the fiber through interactions between these channels, for example, cross-phase modulation (XPM) which is dependent on the relative powers and difference in wavelength between interfering channels.
While the addition or deletion of channels from a line system can be simulated, simulation involves a calculation which is time-consuming and costly to perform within a network element. The simulation is also sensitive to unknown parameters which are difficult to obtain, such as the connector losses which may exist before the input or at the output of the transmission fiber. It is difficult to differentiate this loss from the loss of the fiber itself, and these differences will change the solution which is obtained through simulation. Therefore, in the absence of simulation, the system cannot predict the final state of the channels which will result from any change.
Alternatively, the optimization of a line system can be achieved using a control system which measures a combination of total powers and per-channel powers at various points. Such a controller can be implemented which can iterate to a solution which was not easily calculated by the system, and such a solution will intrinsically include the effect of unknown parameters described in the previous paragraph. A second advantage of this approach is that it will converge even if the calculation of the solution is unknown or inaccurate.
A feedback control loop, e.g., a proportional integral derivative (PID) controller, can be used to optimize the set of channels which may be present at a given time. A typical PID controller may use the optical signal to noise ratio (OSNR) as a metric to optimize the channels while implementing an upper power limit for the channels which is known to limit the non-linear effects to a manageable level.
A problem with this approach is its sensitivity to perturbations which are a natural consequence of channel additions and deletions. Because the line system is essentially non-linear (a perturbation in one channel's power affects other channels), the controller must be limited to small changes to remain stable and converge. Since the controller is iterative, each of these changes must be allowed to converge appropriately before applying another change. The result is that the request to add or delete channels takes time. Although this time can be reduced, there will still be delays when reconfiguring services.
A second problem with this approach is its sensitivity to perturbations which are not intended but can be the result of failures of some part of the system. There are many such events which can happen, such as operator error causing a fiber disconnect, equipment failures causing changes in optical power, fiber cuts where a portion of the transmission fiber is damaged or broken, etc. When these events occur there can be a large change in the number of channels which are transmitted through a particular fiber because of channel adds and drops through ROADM nodes. Although the control system can accommodate this type of change, it will take time to converge to a new optimal condition. The performance of the remaining channels can be compromised during the time it takes for this action to complete.
Other solutions exist which try to keep the spectral power in the EDFAs representative of the full fill spectrum under all conditions. These approaches have worked in the past mainly because of the exclusive use of direct detection, and non-polarization multiplexed signals which were the only optical receiving means available before the commercial introduction of coherent optical receivers. In the presence of these coherent signals, which have been widely adopted in product and by standards bodies, current solutions fail to provide the required system level performance for reasons which will be described in the following sections.
Therefore what is needed is a method and system for transient event stabilization of fiber optic transport systems.